Fig 1.
Study area United Arab Emirates showing highest rainfall event (data from global circulation model).
Table 1.
Collected data from different social media platforms.
Table 2.
Data classified into classes.
Fig 2.
Sample images from final dataset containing classified images based on four classes (a) not relevant, (b) rain, (c) low flood and (d) high flood.
Fig 3.
VGG-16 architecture for converting images into flattened features.
Fig 4.
Sample data from 2,171 rows showing binary coded matrix of text messages and extracted features from images and frames.
Fig 5.
Methodology of the case study from data collection to output.
Table 3.
Model accuracy from different functions in Weka.
Fig 6.
Time series of rainfall depths (a) with frequency of total posts per day, (b) with frequency of images and videos per day.
Table 4.
Different classifier results for model accuracy, Kappa statistics, RMSE, F-measure, Area under Curve (AUC) and Precision Recall Curve (PRC).
Fig 7.
Area under Curve (AUC) for three set of data formats using random forest.
Table 5.
Random forest classifier accuracy for data quality of different social media platforms.